← Back to Sessions

Using Dynamic Time Warping to Assist Conventional Waveform Cross-correlation

Description: 

Waveform cross-correlation is an exceptionally sensitive phase-matched filtering technique that can be harnessed for nuclear explosion monitoring to detect repeating similar seismic events, and to estimate their relative locations and magnitudes by comparison with carefully chosen template waveforms. The method becomes even more powerful when applied to multi-channel geophysical data, i.e., 3-component or array. However, there are outstanding challenges with correlation detectors, most notably a direct dependence on the completeness of the waveform template library that is used to match event signatures in an incoming continuous seismic data stream. In most cases template libraries are limited and slight differences in source location and/or type between template and target events could lead to the non-detection of a signal of interest.

We investigate the value of dynamic time warping as an additional analysis step to make waveform correlation more robust. Dynamic time warping (DTW) analyzes the differences and similarities between two time series and attempts to “unwarp” one time series relative to another by finding the minimum distance between them in a recursive manner, hence DTW can compensate for minor differences between a pair of template and target waveforms to get at the underlying overall similarity. We apply DTW to see if the method will allow us to expand the correlation capability of template libraries and to compensate for their limited sampling. We explore what conditions (e.g., source type, station distance, frequency bands) and/or algorithms generate stronger correlation scores. Several state-of-the-art DTW algorithms are used, and we highlight the successes and limitations of them when applied to synthetic earthquake and observed explosion seismic datasets.

Session: Detecting, Locating, Characterizing and Monitoring Non-earthquake Seismoacoustic Sources [Poster]

Type: Poster

Date: 4/19/2023

Presentation Time: 08:00 AM (local time)

Presenting Author: Marlon D. Ramos

Student Presenter: No

Invited Presentation: 

Authors